The present invention relates generally to an aligning method of a chip, an apparatus and method for inspecting samples using the same aligning method, and a method for manufacturing devices using the same apparatus and method, and specifically to: an aligning method for performing a defect inspection of a device pattern having a pattern of size equal to or less than 0.1 μm formed on a surface of a sample, such as a stencil mask, a wafer and the like, with a high precision, a high reliability and a high resolution, and also with a high throughput; an apparatus and method for inspecting samples using the same aligning method; and a method for manufacturing devices, which includes a step of inspecting samples by using the same apparatus and method for inspecting samples.
An apparatus for inspecting a sample for defects is typically operated in a manner in which an electron beam is irradiated onto a sample to be inspected, such as a wafer, to thereby generate electrons containing data related to a device pattern formed on a sample surface to be inspected; the generated electrons are then used to form an image of the data representing the device pattern; and thus obtained image is inspected in accordance with a predetermined inspection program. To improve the reliability of the result of the inspection, it is required that data having a high precision should be obtained from the device pattern on the sample surface through the irradiation of electrons. One means to address this is represented by the registration of a stage in the X-axis and the Y-axis directions, which carries the sample thereon and moves in the X-axis direction and in the Y-axis direction orthogonal to said X-axis direction, and by a focus adjustment in the Z-axis direction parallel with the axial direction of a secondary optical system.
In the practice according to the prior art method, taking as an example a case where images for two regions corresponding to each other are generated from the wafer surface to be inspected and thus obtained two images, or one of the image and another corresponding image are inspected for any defects, such a method has been typically employed, in which for said one image, a plurality of images is generated each taken by shifting a position by +1 pixel, +2 pixels, −1 pixel, −2 pixels, respectively, along the X-axis direction and the Y-axis direction, and a total of 25 thus obtained images consisting of those 24 shifted images plus 1 not-shifted image are compared with the other images, wherein a defect inspection apparatus using a single electron beam has been put into practical use for forming those images.
Further, an inspection system using a multi-beam to perform a defect inspection of the samples has been also suggested in order to improve the throughput (see, for example, Specification of U.S. Pat. No. 5,892,224 and B. Lischke, Japanese Journal of Applied Physics, Vol. 28, No. 10, p 2058). There is another known method, in which a rectangular beam is irradiated onto a sample, and an electron beam emanated from the irradiated point is magnified by a projection optical system for detection (see, for example, Japanese Patent Laid-open Publication No. Hei7-24939).
Those systems that carry out the defect inspection using a plurality of electron beams with which a plurality of regions can be scanned at once, may be considered to theoretically improve the throughput in proportion to the number of electron beams.
The above-described inspection apparatuses for a pattern defect according to the prior art, however, have been associated with a problem that it could be difficult to perform an accurate inspection, which may possibly be arise for the following reasons:
(1) Although a stage is installed for carrying a sample thereon and moving therewith in the X-axis direction and in the Y-axis direction orthogonal to the X-axis direction, there might be a case where a distortion is induced in a stage guide serving for guiding the stage, or another case where the stage guides in the X-axis and the Y-axis directions are not crossing precisely at a right angle with respect to each other, which would prevent the stage from moving along an ideal track;
(2) Upon placing the sample on the stage, there might be a case where an X-Y coordinate of the sample is not aligned with an X-Y coordinate of the stage, and so an error would be generated in a rotational direction;
(3) There might be a case where an error is introduced in a laser interferometer for detecting a position of the sample;
(4) There might be a case where, in some samples, the die could be formed in a position offset from its designed position in the lithography process;
(5) There might be a case where a variation in moving speed is induced during a continuous movement of the stage; and
(6) There might be a case where a charge-up is induced in the sample by the irradiation of the electron beam and a resultantly obtained image contains a distortion generated therefrom.
If the errors described above are not somehow compensated for, the obtained image could be offset from its theoretical position by ±2 or more pixels, for example. If there is a possibility that said offset occurs to an extent defined by each ±3 pixels in the X-axis and the Y-axis directions, then in order to ensure accurate defect inspection, the number of images to be generated for the comparison should be as much as 7×7=49. Consequently, with the above systems there could be a disadvantageous situation that the number of memories and comparator circuits required for the inspection must be increased, which in turn leads to a problem that the rate of the defect inspection could no longer keep up with that of the image taking, and accordingly the defect inspection could not be performed with high throughput.
In addition, in the defect inspection of the samples according to the prior art, as described above, simply the registration in the X-Y directions is typically practiced prior to two-dimensional image taking for subsequent pattern inspection, but an uneven surface of the sample has not been taken into consideration. From this reason also, it is possible that highly accurate image signals will not be obtained.
For example, a defect inspection apparatus using a projection optical system, which is known as an apparatus for obtaining a two-dimensional image for inspecting a sample or the like for any defects in the sample, has been associated with a problem that a magnification of a secondary electron image varies significantly over time or in response to any changes in the environment, such as a temperature change. Further, such a projection optical system has another problem that if the surface of the sample is uneven, a resolution of the two-dimensional image deteriorates because of a shallow focal depth of the system.
Yet further, those defect inspection apparatuses according to the prior art have been associated with a problem that an accurate defect inspection can not be carried out due to a frequent variation in the magnification of the image projection optical system, and in addition, no special attention has been paid to a need for an accurate measuring of a scanning sensibility of a multi-beam optical system, and also no reference specifically disclosing this matter has been found.
Still further, for the SEM using a single beam, since it comprises a single beam and a single detector, and accordingly the density of the signal fully represents data on the sample, therefore the defect detection can be performed by simply carrying out the pattern matching, but for the case of using multi beams, since the multi beams contain the variation in its beam current value by some percentage among respective beams and also has a difference in the detecting sensibility among respective beams. Therefore, the density of the signal is not necessarily representing the data on the sample exclusively. Besides, the defect detection method using the projection optical system has a problem that the density of the signal could be different even for the same pattern section in a sample depending on whether it is located in the marginal area of the field of view or in the area adjacent to an optical axis, and this may lead to a frequent detection of false-defect during the defect detecting operation.